This repository provides a comprehensive survey of research papers focused on Personal LLM Agents, which are LLM-based agents designed for deep integration with personal data, devices, and services, often targeting resource-constrained mobile or edge environments. It serves researchers and developers in the field by cataloging advancements in capabilities, efficiency, and security.

How It Works

The survey categorizes papers across key aspects of Personal LLM Agents: Task Automation (especially UI-grounded), Sensing (user activity, environment), Memorization (obtaining, managing, self-evolution), Efficiency (inference, memory retrieval), and Security/Privacy (confidentiality, integrity, reliability). It highlights both LLM-based and traditional approaches within these categories, offering a structured overview of the research landscape.

Quick Start & Requirements

This repository is a curated list of academic papers and does not involve code execution or installation. Links to papers, code repositories, and discussion forums are provided within the README.

Highlighted Details

• Extensive coverage of UI-grounded task automation, including numerous LLM-based and traditional methods.
• Detailed sections on sensing techniques for context awareness and memory management strategies for personalization.
• Focus on efficiency aspects like vector databases and approximate nearest neighbor search for memory optimization.
• Comprehensive breakdown of security and privacy concerns, including adversarial attacks and data confidentiality.

Maintenance & Community

The survey is associated with a paper published on arXiv and includes a link to a Zulip discussion group for community engagement. The authors acknowledge feedback from numerous industry experts.

Licensing & Compatibility

The repository itself is a collection of links and does not have a specific license. Individual papers retain their original licensing.

Limitations & Caveats

This repository is a survey and does not provide executable code or benchmarks. The rapid evolution of LLM research means that new papers and advancements may not be immediately reflected.